SAS Viya

Overview

SAS Viya is an integrated software platform designed for advanced analytics, artificial intelligence, and machine learning workloads within enterprise environments. It provides a unified architecture for the entire analytical lifecycle, from data ingestion and preparation to model development, deployment, and monitoring. The platform is engineered to support a range of users, including data scientists, business analysts, and developers, by offering various interfaces and tools tailored to different skill sets and use cases.

At its core, SAS Viya leverages in-memory processing through its Cloud Analytic Services (CAS) engine, which is designed to handle large datasets and complex computations efficiently. This architecture enables users to perform sophisticated statistical analysis, build and train machine learning models, develop forecasting solutions, and conduct text analytics at scale. The platform supports a multi-language approach, allowing users to interact with its capabilities using SAS, Python, R, Java, and Lua, facilitating integration into existing data science workflows and codebases.

SAS Viya's primary utility lies in its ability to provide an end-to-end solution for organizations seeking to operationalize AI and machine learning. It addresses challenges associated with model governance, reproducibility, and scalability, which are critical in regulated industries and large-scale deployments. The platform includes tools for automated machine learning (AutoML), explainable AI (XAI), and model management, which assist in streamlining the development process and ensuring transparency and control over deployed models. For example, SAS Model Manager assists in monitoring model performance and detecting drift post-deployment through its documentation.

The platform is suitable for organizations requiring robust security, compliance, and auditing capabilities, given its adherence to standards like SOC 2 Type II, ISO 27001, GDPR, and HIPAA. These compliance certifications are particularly relevant for enterprises operating in finance, healthcare, and government sectors. Its comprehensive suite of products, such as SAS Visual Analytics for interactive data exploration and SAS Visual Data Mining and Machine Learning for model building, collectively aim to support diverse analytical needs within a consistent and governed environment.

While SAS Viya offers a broad range of features, its enterprise focus means it is typically deployed in scenarios where data volume, analytical complexity, and regulatory requirements are high. Alternatives like Databricks also provide unified platforms for data and AI, often emphasizing cloud-native, open-source ecosystems. The choice between such platforms often depends on an organization's existing technology stack, preference for proprietary versus open-source tools, and specific governance needs.

Key features

• Cloud Analytic Services (CAS): An in-memory, distributed processing engine designed for high-performance analytics on large datasets.
• Multi-language Support: Enables users to write code in SAS, Python, R, Java, and Lua, integrating with various developer ecosystems.
• Visual Interfaces: Provides drag-and-drop interfaces for data exploration, model building, and report generation, catering to users without extensive coding experience.
• Automated Machine Learning (AutoML): Tools to automate model selection, hyperparameter tuning, and feature engineering, accelerating model development.
• Model Management and Governance: Capabilities for versioning, deploying, monitoring, and retraining models, ensuring compliance and performance over time.
• Explainable AI (XAI): Features to interpret model predictions and understand underlying factors, enhancing trust and transparency in AI systems.
• Text Analytics and Natural Language Processing (NLP): Tools to extract insights from unstructured text data, including sentiment analysis and topic modeling.
• Forecasting and Time Series Analysis: Specialized algorithms and tools for developing predictive models based on historical time-series data.
• Data Preparation and Transformation: Capabilities for cleaning, transforming, and integrating data from various sources to prepare it for analysis.

Pricing

SAS Viya utilizes a custom enterprise pricing model. Specific costs are determined based on an organization's unique requirements, including the scale of deployment, the number of users, and the specific modules and capabilities required. Interested parties must contact SAS directly for a personalized quote.

Common integrations

• Python: Integration through SAS Scripting Wrapper for Analytics Transfer (SWAT) package, enabling Python users to interact with CAS data and procedures as described in SAS documentation.
• R: Similar to Python, R users can connect to and leverage CAS capabilities for scalable analytics.
• Java: APIs and client libraries for integrating SAS Viya services into Java applications.
• REST APIs: Extensive RESTful API ecosystem for programmatic access to SAS Viya services, data, and models.
• Apache Spark: Connectivity to Spark clusters for data ingestion and processing, allowing for hybrid analytical workflows.
• Cloud Platforms: Deployment options and integrations with major cloud providers like AWS, Azure, and Google Cloud Platform.
• Data Visualization Tools: Export capabilities and direct connections to external visualization tools for custom reporting.

Alternatives

• Databricks: A unified data and AI platform built on Apache Spark, offering data warehousing, machine learning, and data engineering capabilities in a cloud-native environment.
• H2O.ai: Provides open-source and commercial AI platforms, including H2O-3 and H2O Driverless AI, focusing on automated machine learning and MLOps.
• Alteryx: A platform for data science and analytics that emphasizes self-service data preparation, blending, and advanced analytics with a low-code/no-code interface.

Getting started

To interact with SAS Viya programmatically, especially for data scientists, using the Python SWAT package is a common approach. This package allows Python users to connect to a SAS Cloud Analytic Services (CAS) server, load data, and execute CAS actions.

First, ensure the sas-swat package is installed:

pip install sas-swat

Then, you can establish a connection and perform a simple data loading operation:

import swat

# Replace with your CAS server details
cas_host = 'your-cas-server.example.com'
cas_port = 5570
cas_username = 'your_username'
cas_password = 'your_password'

# Connect to the CAS server
conn = swat.CAS(hostname=cas_host, port=cas_port, username=cas_username, password=cas_password)
print(f"Connected to CAS server: {conn.serverstatus().cas.hostname}")

# Example: Create a simple DataFrame and load it to CAS
import pandas as pd
data = pd.DataFrame({
    'id': [1, 2, 3],
    'value': [10, 20, 30]
})

# Load the DataFrame to a CAS table
cas_table_name = 'my_test_table'
cas_table = conn.upload_frame(data, casout={'name': cas_table_name, 'replace': True})
print(f"DataFrame loaded to CAS table: {cas_table_name}")

# Verify the table exists and get its information
print(cas_table.tableinfo())

# Perform a simple action, e.g., get data statistics
result = conn.simple.summary(table=cas_table_name)
print("Summary statistics:")
print(result)

# Disconnect from CAS
conn.close()
print("Disconnected from CAS server.")

This Python script demonstrates connecting to a CAS server, uploading a Pandas DataFrame, and running a basic summary statistics action. This workflow is foundational for leveraging SAS Viya's analytical capabilities from a Python environment.